Clean up netstack README and TODO situation.

examples/netstack/README.md

@@ -1,5 +1,18 @@
 # TCP/IP Stack
 
+There are two (closely-related) implementations here:
+
+ - [`monolithic-lowlevel`](monolithic-lowlevel/) is the original
+   implementation, originally written for `minimart`, a language that
+   followed our ESOP 2014 paper quite closely. Porting it to a
+   monolithic-assertion-set Syndicate dialect helped substantially
+   simplify the code.
+
+ - [`incremental-highlevel`](incremental-highlevel/) is a port of
+   `monolithic-lowlevel` to the Syndicate high-level DSL
+   ("`syndicate/actor`"). Moving from the low-level Syndicate style to
+   the high-level style also drastically simplified the code.
+
 ## Linux Firewall Configuration
 
 Imagine a setup where the machine you are running this code has IP

examples/netstack/TODO.md

@@ -11,8 +11,14 @@ TCP options negotiation.
  - SACK
  - Window scaling
 
+Check that we handle the situations in figs. 9, 10, 11, pp.33- of RFC 793.
+
 Bugs:
  - RST kills a connection even if its sequence number is bogus. Check
    to make sure it's in the window. (See
    http://static.googleusercontent.com/media/research.google.com/en//pubs/archive/41848.pdf
    and RFC 5961)
+
+   Conform better to the rules for reset generation and processing
+   from pp.36- of RFC 793. In particular, do not blindly accept RSTs
+   without checking sequence numbers against windows etc.

examples/netstack/monolithic-lowlevel/README.md

@@ -1,16 +0,0 @@
-# TCP/IP Stack
-
-## Linux Firewall Configuration
-
-Imagine a setup where the machine you are running this code has IP
-192.168.1.10. This code claims 192.168.1.222 for itself. Now, pinging
-192.168.1.222 from some other machine, say 192.168.1.99, will cause
-the local kernel to receive the pings and then *forward them on to
-192.168.1.222*, which because of the gratuitous ARP announcement, it
-knows to be on its own Ethernet MAC address. This causes the ping
-requests to repeat endlessly, each time with one lower TTL.
-
-One approach to solving the problem is to prevent the kernel from
-forwarding packets addressed to 192.168.1.222. To do this,
-
-    sudo iptables -I FORWARD -d 192.168.1.222 -j DROP

examples/netstack/monolithic-lowlevel/TODO.md

@@ -1,18 +0,0 @@
-Ideas on TCP unit testing:
-<https://www.snellman.net/blog/archive/2015-07-09-unit-testing-a-tcp-stack/>
-
-Check behaviour around TCP zero-window probing. Is the correct
-behaviour already a consequence of the way `send-outbound` works?
-
-Do something smarter with TCP timers and RTT estimation than the
-nothing that's already being done.
-
-TCP options negotiation.
- - SACK
- - Window scaling
-
-Bugs:
- - RST kills a connection even if its sequence number is bogus. Check
-   to make sure it's in the window. (See
-   http://static.googleusercontent.com/media/research.google.com/en//pubs/archive/41848.pdf
-   and RFC 5961)